icuSources/i18n/regexcmp.h

   1 //
   2 //  regexcmp.h
   3 //
   4 //  Copyright (C) 2002-2015, International Business Machines Corporation and others.
   5 //  All Rights Reserved.
   6 //
   7 //  This file contains declarations for the class RegexCompile
   8 //
   9 //  This class is internal to the regular expression implementation.
  10 //  For the public Regular Expression API, see the file "unicode/regex.h"
  11 //
  12
  13
  14 #ifndef RBBISCAN_H
  15 #define RBBISCAN_H
  16
  17 #include "unicode/utypes.h"
  18 #if !UCONFIG_NO_REGULAR_EXPRESSIONS
  19
  20 #include "unicode/uobject.h"
  21 #include "unicode/uniset.h"
  22 #include "unicode/parseerr.h"
  23 #include "uhash.h"
  24 #include "uvector.h"
  25 #include "uvectr32.h"
  26
  27
  28
  29 U_NAMESPACE_BEGIN
  30
  31
  32 //--------------------------------------------------------------------------------
  33 //
  34 //  class RegexCompile    Contains the regular expression compiler.
  35 //
  36 //--------------------------------------------------------------------------------
  37 struct  RegexTableEl;
  38 class   RegexPattern;
  39
  40
  41 class U_I18N_API RegexCompile : public UMemory {
  42 public:
  43
  44     enum {
  45         kStackSize = 100            // The size of the state stack for
  46     };                              //   pattern parsing.  Corresponds roughly
  47                                     //   to the depth of parentheses nesting
  48                                     //   that is allowed in the rules.
  49
  50     struct RegexPatternChar {
  51         UChar32             fChar;
  52         UBool               fQuoted;
  53     };
  54
  55     RegexCompile(RegexPattern *rp, UErrorCode &e);
  56
  57     void       compile(const UnicodeString &pat, UParseError &pp, UErrorCode &e);
  58     void       compile(UText *pat, UParseError &pp, UErrorCode &e);
  59
  60
  61     virtual    ~RegexCompile();
  62
  63     void        nextChar(RegexPatternChar &c);      // Get the next char from the input stream.
  64
  65     static void cleanup();                       // Memory cleanup
  66
  67
  68
  69     // Categories of parentheses in pattern.
  70     //   The category is saved in the compile-time parentheses stack frame, and
  71     //   determines the code to be generated when the matching close ) is encountered.
  72     enum EParenClass {
  73         plain        = -1,               // No special handling
  74         capturing    = -2,
  75         atomic       = -3,
  76         lookAhead    = -4,
  77         negLookAhead = -5,
  78         flags        = -6,
  79         lookBehind   = -7,
  80         lookBehindN  = -8
  81     };
  82
  83 private:
  84
  85
  86     UBool       doParseActions(int32_t a);
  87     void        error(UErrorCode e);                   // error reporting convenience function.
  88
  89     UChar32     nextCharLL();
  90     UChar32     peekCharLL();
  91     UnicodeSet  *scanProp();
  92     UnicodeSet  *scanPosixProp();
  93     void        handleCloseParen();
  94     int32_t     blockTopLoc(UBool reserve);          // Locate a position in the compiled pattern
  95                                                      //  at the top of the just completed block
  96                                                      //  or operation, and optionally ensure that
  97                                                      //  there is space to add an opcode there.
  98     void        compileSet(UnicodeSet *theSet);      // Generate the compiled pattern for
  99                                                      //   a reference to a UnicodeSet.
 100     void        compileInterval(int32_t InitOp,      // Generate the code for a {min,max} quantifier.
 101                                int32_t LoopOp);
 102     UBool       compileInlineInterval();             // Generate inline code for a {min,max} quantifier
 103     void        literalChar(UChar32 c);              // Compile a literal char
 104     void        fixLiterals(UBool split=FALSE);      // Generate code for pending literal characters.
 105     void        insertOp(int32_t where);             // Open up a slot for a new op in the
 106                                                      //   generated code at the specified location.
 107     void        appendOp(int32_t op);                // Append a new op to the compiled pattern.
 108     void        appendOp(int32_t type, int32_t val); // Build & append a new op to the compiled pattern.
 109     int32_t     buildOp(int32_t type, int32_t val);  // Construct a new pcode instruction.
 110     int32_t     allocateData(int32_t size);          // Allocate space in the matcher data area.
 111                                                      //   Return index of the newly allocated data.
 112     int32_t     allocateStackData(int32_t size);     // Allocate space in the match back-track stack frame.
 113                                                      //   Return offset index in the frame.
 114     int32_t     minMatchLength(int32_t start,
 115                                int32_t end);
 116     int32_t     maxMatchLength(int32_t start,
 117                                int32_t end);
 118     void        matchStartType();
 119     void        stripNOPs();
 120
 121     void        setEval(int32_t op);
 122     void        setPushOp(int32_t op);
 123     UChar32     scanNamedChar();
 124     UnicodeSet *createSetForProperty(const UnicodeString &propName, UBool negated);
 125
 126 public:   // Public for testing only.
 127     static void U_EXPORT2 findCaseInsensitiveStarters(UChar32 c, UnicodeSet *starterChars);
 128 private:
 129
 130
 131     UErrorCode                    *fStatus;
 132     RegexPattern                  *fRXPat;
 133     UParseError                   *fParseErr;
 134
 135     //
 136     //  Data associated with low level character scanning
 137     //
 138     int64_t                       fScanIndex;        // Index of current character being processed
 139                                                      //   in the rule input string.
 140     UBool                         fQuoteMode;        // Scan is in a \Q...\E quoted region
 141     UBool                         fInBackslashQuote; // Scan is between a '\' and the following char.
 142     UBool                         fEOLComments;      // When scan is just after '(?',  inhibit #... to
 143                                                      //   end of line comments, in favor of (?#...) comments.
 144     int64_t                       fLineNum;          // Line number in input file.
 145     int64_t                       fCharNum;          // Char position within the line.
 146     UChar32                       fLastChar;         // Previous char, needed to count CR-LF
 147                                                      //   as a single line, not two.
 148     UChar32                       fPeekChar;         // Saved char, if we've scanned ahead.
 149
 150
 151     RegexPatternChar              fC;                // Current char for parse state machine
 152                                                      //   processing.
 153
 154     //
 155     //   Data for the state machine that parses the regular expression.
 156     //
 157     RegexTableEl                  **fStateTable;     // State Transition Table for regex Rule
 158                                                      //   parsing.  index by p[state][char-class]
 159
 160     uint16_t                      fStack[kStackSize];  // State stack, holds state pushes
 161     int32_t                       fStackPtr;           //  and pops as specified in the state
 162                                                        //  transition rules.
 163
 164     //
 165     //  Data associated with the generation of the pcode for the match engine
 166     //
 167     int32_t                       fModeFlags;        // Match Flags.  (Case Insensitive, etc.)
 168                                                      //   Always has high bit (31) set so that flag values
 169                                                      //   on the paren stack are distinguished from relocatable
 170                                                      //   pcode addresses.
 171     int32_t                       fNewModeFlags;     // New flags, while compiling (?i, holds state
 172                                                      //   until last flag is scanned.
 173     UBool                         fSetModeFlag;      // true for (?ismx, false for (?-ismx
 174
 175     UnicodeString                 fLiteralChars;     // Literal chars or strings from the pattern are accumulated here.
 176                                                      //   Once completed, meaning that some non-literal pattern
 177                                                      //   construct is encountered, the appropriate opcodes
 178                                                      //   to match the literal will be generated, and this
 179                                                      //   string will be cleared.
 180
 181     int64_t                       fPatternLength;    // Length of the input pattern string.
 182
 183     UVector32                     fParenStack;       // parentheses stack.  Each frame consists of
 184                                                      //   the positions of compiled pattern operations
 185                                                      //   needing fixup, followed by negative value.  The
 186                                                      //   first entry in each frame is the position of the
 187                                                      //   spot reserved for use when a quantifier
 188                                                      //   needs to add a SAVE at the start of a (block)
 189                                                      //   The negative value (-1, -2,...) indicates
 190                                                      //   the kind of paren that opened the frame.  Some
 191                                                      //   need special handling on close.
 192
 193
 194     int32_t                       fMatchOpenParen;   // The position in the compiled pattern
 195                                                      //   of the slot reserved for a state save
 196                                                      //   at the start of the most recently processed
 197                                                      //   parenthesized block. Updated when processing
 198                                                      //   a close to the location for the corresponding open.
 199
 200     int32_t                       fMatchCloseParen;  // The position in the pattern of the first
 201                                                      //   location after the most recently processed
 202                                                      //   parenthesized block.
 203
 204     int32_t                       fIntervalLow;      // {lower, upper} interval quantifier values.
 205     int32_t                       fIntervalUpper;    // Placed here temporarily, when pattern is
 206                                                      //   initially scanned.  Each new interval
 207                                                      //   encountered overwrites these values.
 208                                                      //   -1 for the upper interval value means none
 209                                                      //   was specified (unlimited occurences.)
 210
 211     int64_t                       fNameStartPos;     // Starting position of a \N{NAME} name in a
 212                                                      //   pattern, valid while remainder of name is
 213                                                      //   scanned.
 214
 215     UStack                        fSetStack;         // Stack of UnicodeSets, used while evaluating
 216                                                      //   (at compile time) set expressions within
 217                                                      //   the pattern.
 218     UStack                        fSetOpStack;       // Stack of pending set operators (&&, --, union)
 219
 220     UChar32                       fLastSetLiteral;   // The last single code point added to a set.
 221                                                      //   needed when "-y" is scanned, and we need
 222                                                      //   to turn "x-y" into a range.
 223
 224     UnicodeString                *fCaptureName;      // Named Capture, the group name is built up
 225                                                      //   in this string while being scanned.
 226 };
 227
 228 // Constant values to be pushed onto fSetOpStack while scanning & evalueating [set expressions]
 229 //   The high 16 bits are the operator precedence, and the low 16 are a code for the operation itself.
 230
 231 enum SetOperations {
 232     setStart         = 0 << 16 | 1,
 233     setEnd           = 1 << 16 | 2,
 234     setNegation      = 2 << 16 | 3,
 235     setCaseClose     = 2 << 16 | 9,
 236     setDifference2   = 3 << 16 | 4,    // '--' set difference operator
 237     setIntersection2 = 3 << 16 | 5,    // '&&' set intersection operator
 238     setUnion         = 4 << 16 | 6,    // implicit union of adjacent items
 239     setDifference1   = 4 << 16 | 7,    // '-', single dash difference op, for compatibility with old UnicodeSet.
 240     setIntersection1 = 4 << 16 | 8     // '&', single amp intersection op, for compatibility with old UnicodeSet.
 241     };
 242
 243 U_NAMESPACE_END
 244 #endif   // !UCONFIG_NO_REGULAR_EXPRESSIONS
 245 #endif   // RBBISCAN_H